A LIDAR LOCALIZATION METHOD BASED ON SPATIO-TEMPORAL FUSION AND QUALITY FILTERING

Vehicle localization is one of the primary challenges in autonomous driving. LiDAR, due to its wide detection range and high distance accuracy, has been widely applied in the localization tasks of autonomous driving. Traditional LiDAR localization algorithms rely solely on the pose obtained from matching the current single-frame point cloud. However, due to the sparsity of point cloud, single-frame matching methods struggle to avoid localization errors.To solve this problem, this paper proposes a LiDAR localization method based on spatiotemporal fusion and quality filtering. Firstly, a spatiotemporal fused pose set is constructed to take advantage of spatiotemporal connectivity between LiDAR data. Then, a quality filtering process is applied to select the best poses from the pose set. Finally, the best poses are further optimized to improve the localization accuracy. The performance of the proposed method is evaluated using both open-source data and real-world measured data, validating the effectiveness of the proposed approach.